Signal modulation identification method, device, equipment and storage medium

Abstract

The present application provides a signal modulation identification method, device, equipment and storage medium, wherein the method includes: obtaining a signal to be identified; generating an equipotential star map corresponding to the signal to be identified according to the signal to be identified; inputting the equipotential star map into a first signal identification model for initial identification to obtain an initial signal category of the signal to be identified; judging the signal category of the signal to be identified, and if the signal category of the signal to be identified is multi-base orthogonal amplitude modulation, determining the target signal type of the signal to be identified, and performing secondary identification on the signal to be identified based on a second signal identification model corresponding to the target signal type, to obtain an identification result of the signal to be identified. The present application can ensure the accuracy of signal modulation identification, and at the same time, improve the generalization ability of signal identification, and, due to the secondary identification, also enhance the anti-interference ability of signal modulation identification, and realize the rapid and accurate identification of digital modulated signals.

Description

Technical Field

[0001] The present application relates to the field of signal processing technology, and in particular to a signal modulation identification method, device, equipment and storage medium. Background Art

[0002] Automatic Modulation Recognition (AMR) can detect the modulation mode of the received signal, which facilitates the subsequent signal demodulation, spectrum monitoring and signal processing. At the same time, it is also an indispensable part of signal interception and analysis, and plays an important role in wireless communications, spectrum management and radar systems.

[0003] In the prior art, an AMR method based on deep learning can be adopted, using deep learning technologies such as convolutional neural networks (CNN), recurrent neural networks (RNN), long short-term memory networks (LSTM), Transformer, etc., to achieve adaptive recognition of modulated signals.

[0004] However, for digital modulation, since there are many signal categories to be identified and some modulation categories have high orders, the deep learning methods in the prior art have the problem of low recognition accuracy when dealing with recognition tasks of multiple modulation categories. Summary of the invention

[0005] The purpose of this application is to provide a signal modulation recognition method, device, equipment and storage medium to address the deficiencies in the above-mentioned prior art, so as to solve the problem of low recognition accuracy in the prior art.

[0006] To achieve the above purpose, the technical solution adopted in the embodiment of the present application is as follows:

[0007] In a first aspect, an embodiment of the present application provides a signal modulation identification method, the method comprising:

[0008] Acquire a signal to be identified;

[0009] According to the signal to be identified, generating an equipotential planet diagram corresponding to the signal to be identified;

[0010] Inputting the equipotential star map into a first signal recognition model for initial recognition to obtain an initial signal category of the signal to be recognized, wherein the first signal recognition model is a MobileNet V2 model trained using signal data of multiple categories as samples;

[0011] The signal category of the signal to be identified is judged. If the signal category of the signal to be identified is multi-base quadrature amplitude modulation, the target signal type of the signal to be identified is determined, and the signal to be identified is re-identified based on a second signal recognition model corresponding to the target signal type to obtain a recognition result of the signal to be identified, where the second signal recognition model is a MobileNet V2 model trained using signal data having the target signal type as samples.

[0012] In a possible implementation manner, determining the target signal type of the signal to be identified includes:

[0013] The target signal type of the signal to be identified is determined based on the high-order cumulative amount characteristics of the signal to be identified, and the target signal type includes: square quadrature amplitude modulation or cross quadrature amplitude modulation.

[0014] In a possible implementation, performing secondary recognition on the signal to be recognized based on the second signal recognition model corresponding to the target signal type to obtain a recognition result of the signal to be recognized includes:

[0015] Determining a normalized instantaneous amplitude characteristic of the signal to be identified;

[0016] The normalized instantaneous amplitude feature is input into a second signal recognition model corresponding to the target signal type for secondary recognition to obtain a recognition result of the signal to be recognized.

[0017] In a possible implementation, the method further includes:

[0018] If the signal category of the signal to be identified is multi-level frequency shift keying, Fourier transform is performed on the signal to be identified to obtain the transformed signal to be identified, high-order cumulative quantity characteristics of the transformed signal to be identified are calculated, and the identification result of the signal to be identified is determined based on the high-order cumulative quantity characteristics.

[0019] In a possible implementation, the method further includes:

[0020] If the signal category of the signal to be identified is a category other than multi-level quadrature amplitude modulation and multi-level frequency shift keying, the initial signal category is used as the identification result of the signal to be identified.

[0021] In a possible implementation, before generating an equipotential star map corresponding to the signal to be identified according to the signal to be identified, the method includes:

[0022] It is determined whether the signal to be identified is an in-phase orthogonal signal; if not, the signal to be identified is converted into an in-phase orthogonal signal.

[0023] In a possible implementation, the step of inputting the equipotential star map into a first signal recognition model for initial recognition to obtain an initial signal category of the signal to be recognized includes:

[0024] Loading a weight file of the first signal recognition model, wherein the weight file is used to indicate parameters in the first signal recognition model;

[0025] Loading the weight file into the first signal recognition model;

[0026] The equipotential star map is input into the first signal recognition model for initial recognition to obtain an initial signal category of the signal to be recognized.

[0027] In a second aspect, another embodiment of the present application provides a signal modulation identification device, the device comprising:

[0028] An acquisition module, used for acquiring a signal to be identified;

[0029] A generating module, used for generating an equipotential star map corresponding to the signal to be identified according to the signal to be identified;

[0030] A primary recognition module, used for inputting the equipotential star map into a first signal recognition model for primary recognition to obtain an initial signal category of the signal to be recognized, wherein the first signal recognition model is a MobileNet V2 model trained with signal data of multiple categories as samples;

[0031] A secondary recognition module is used to judge the signal category of the signal to be recognized. If the signal category of the signal to be recognized is multi-base quadrature amplitude modulation, the target signal type of the signal to be recognized is determined, and the signal to be recognized is secondary recognized based on a second signal recognition model corresponding to the target signal type to obtain a recognition result of the signal to be recognized, where the second signal recognition model is a MobileNet V2 model trained using signal data having the target signal type as a sample.

[0032] In a possible implementation, the secondary identification module is specifically used to:

[0033] The target signal type of the signal to be identified is determined based on the high-order cumulative amount characteristics of the signal to be identified, and the target signal type includes: square quadrature amplitude modulation or cross quadrature amplitude modulation.

[0034] In a possible implementation, the secondary identification module is specifically used to:

[0035] Determining a normalized instantaneous amplitude characteristic of the signal to be identified;

[0036] The normalized instantaneous amplitude feature is input into a second signal recognition model corresponding to the target signal type for secondary recognition to obtain a recognition result of the signal to be recognized.

[0037] In a possible implementation, the device further includes: a determination module;

[0038] The determination module is used to perform Fourier transform on the signal to be identified if the signal category of the signal to be identified is multi-level frequency shift keying, obtain the transformed signal to be identified, calculate the high-order cumulative amount characteristics of the transformed signal to be identified, and determine the identification result of the signal to be identified based on the high-order cumulative amount characteristics.

[0039] In a possible implementation, the determining module is further configured to:

[0040] If the signal category of the signal to be identified is a category other than multi-level quadrature amplitude modulation and multi-level frequency shift keying, the initial signal category is used as the identification result of the signal to be identified.

[0041] In a possible implementation, the generating module is further used to:

[0042] It is determined whether the signal to be identified is an in-phase orthogonal signal; if not, the signal to be identified is converted into an in-phase orthogonal signal.

[0043] In a possible implementation, the initial identification module is specifically used to:

[0044] Loading a weight file of the first signal recognition model, wherein the weight file is used to indicate parameters in the first signal recognition model;

[0045] Loading the weight file into the first signal recognition model;

[0046] The equipotential star map is input into the first signal recognition model for initial recognition to obtain an initial signal category of the signal to be recognized.

[0047] In the third aspect, another embodiment of the present application provides an electronic device, comprising: a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium through the bus, and the processor executes the machine-readable instructions to perform the steps of any method described in the first aspect above.

[0048] In a fourth aspect, another embodiment of the present application provides a storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any method described in the first aspect are executed.

[0049] The beneficial effects of the present application are as follows: by acquiring a signal to be identified; and generating an equipotential star diagram corresponding to the signal to be identified based on the signal to be identified, the symbol characteristics of the signal to be identified in different modulation modes can be characterized, so that the equipotential star diagram is input into the first signal recognition model for initial recognition, and the initial signal category of the signal to be identified is obtained, and the category of the signal to be identified is determined, and the initial signal category of the signal to be identified is obtained. On this basis, by judging the signal category of the signal to be identified, when the signal category of the signal to be identified is multi-level orthogonal amplitude modulation, the target signal type of the signal to be identified is determined, and based on the second signal recognition model corresponding to the target signal type, the signal to be identified is secondary identified, and the identification result of the signal to be identified is obtained. On the basis of the primary identification, the order of the signal to be identified in the orthogonal amplitude modulation can be determined, and the secondary identification of the multi-level orthogonal amplitude modulation can be realized. That is to say, when identifying signal modulation in the present application, first, the signal to be identified is identified once by a first signal recognition model obtained by training with signal data of multiple categories as samples, and the modulation method adopted by the signal to be identified is classified once to determine the category to which the modulation method of the signal to be identified belongs. Moreover, on the basis of the first recognition, the signal to be identified is secondarily identified by a second signal recognition model obtained by training with signal data of the target signal type as samples to determine the specific modulation method of the signal to be identified. Moreover, the signal data of multiple categories are different from the signal data of the target signal type, which can ensure the accuracy of signal modulation recognition and improve the generalization ability of signal recognition. In addition, due to the secondary recognition, the anti-interference ability of signal modulation recognition is also enhanced, thereby realizing the rapid and accurate recognition of digital modulated signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 A schematic diagram of a flow chart of a signal modulation identification method provided in an embodiment of the present application;

[0052] Figure 2 A schematic diagram of a flow chart for obtaining a recognition result of a signal to be recognized in the signal modulation recognition method provided in an embodiment of the present application;

[0053] Figure 3 A schematic diagram of a flow chart for obtaining an initial signal category of a signal to be identified in the signal modulation identification method provided in an embodiment of the present application;

[0054] Figure 4 A schematic diagram of the structure of the bottleneck layer in the MobileNet V2 model in the signal modulation recognition method provided in an embodiment of the present application;

[0055] Figure 5 An equipotential star diagram of part of the signal in the signal modulation identification method provided in an embodiment of the present application;

[0056] Figure 6 A schematic diagram of recognition accuracy curves under different signal-to-noise ratios for six modulation modes in the signal modulation recognition method provided in an embodiment of the present application;

[0057] Figure 7 A schematic diagram of confusion matrices of six modulation modes in the signal modulation identification method provided in an embodiment of the present application;

[0058] Figure 8 A schematic diagram of the recognition accuracy curve under different signal-to-noise ratios of the square QAM in the signal modulation recognition method provided in an embodiment of the present application;

[0059] Fig. 9 A schematic diagram of a confusion matrix of a square QAM in a signal modulation identification method provided in an embodiment of the present application;

[0060] Fig.10 A schematic diagram of the recognition accuracy curve under different signal-to-noise ratios of the cross QAM in the signal modulation recognition method provided in an embodiment of the present application;

[0061] Fig.11 A schematic diagram of a confusion matrix of a cross QAM in a signal modulation identification method provided in an embodiment of the present application;

[0062] Fig.12 A schematic diagram of a signal modulation identification device provided in an embodiment of the present application;

[0063] Fig.13 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0064] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart under the guidance of the content of the present application, or remove one or more operations from the flowchart.

[0065] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0066] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.

[0067] In the prior art, an AMR method based on deep learning can be adopted, using deep learning technologies such as convolutional neural networks (CNN), recurrent neural networks (RNN), long short-term memory networks (LSTM), Transformer, etc., to achieve adaptive recognition of modulated signals.

[0068] However, for digital modulation, since there are many signal categories to be identified and some modulation categories have high orders, the deep learning methods in the prior art have the problem of low recognition accuracy when dealing with recognition tasks of multiple modulation categories.

[0069] Based on the above problems, the embodiment of the present application proposes a signal modulation identification method, which obtains the signal to be identified; and generates an equipotential star diagram corresponding to the signal to be identified according to the signal to be identified; thereby inputting the equipotential star diagram into the first signal recognition model for initial recognition to obtain the initial signal category of the signal to be identified; and judging the signal category of the signal to be identified, if the signal category of the signal to be identified is multi-base orthogonal amplitude modulation, then determining the target signal type of the signal to be identified, and performing secondary recognition on the signal to be identified based on the second signal recognition model corresponding to the target signal type, to obtain the recognition result of the signal to be identified. It can improve the accuracy, generalization ability and anti-interference ability of signal modulation recognition, and realize the rapid recognition of digital modulation signals.

[0070] The signal modulation identification method provided in the embodiments of the present application is described in detail below in combination with multiple embodiments.

[0071] Figure 1 A flow chart of a signal modulation identification method provided in an embodiment of the present application, referring to Figure 1 As shown, the execution subject of the method can be any electronic device with processing capability, and the method includes:

[0072] S101: Acquire a signal to be identified.

[0073] It can be understood that when identifying the modulation mode of a signal, it is necessary to obtain the signal to be identified.

[0074] Among them, the modulation type of the signal to be identified may include: Multiple Quadrature Amplitude Modulation (MQAM), Binary Phase Shift Keying (BPSK), Quadrature Phase Shift Keying (QPSK), 8-Phase Shift Keying (8PSK) and Multiple Frequency Shift Keying (MFSK).

[0075] S102: Generate an equipotential planet diagram corresponding to the signal to be identified according to the signal to be identified.

[0076] Optionally, after obtaining the signal to be identified, the signal to be identified can be preprocessed to reduce redundant information in the signal to be identified, and the frequency, amplitude and parameters of the signal to be identified can be analyzed to generate an equipotential planet diagram corresponding to the signal to be identified. The equipotential planet diagram is used to describe the mapping of the symbol points of the signal to be identified on the complex plane.

[0077] Exemplarily, the signal to be identified can be used to generate an equipotential planet map corresponding to the signal to be identified through a script, wherein the script can be a program code running in MATLAB software.

[0078] S103, inputting the equipotential star map into the first signal recognition model for initial recognition to obtain an initial signal category of the signal to be recognized.

[0079] Among them, the first signal recognition model is a MobileNet V2 model trained using signal data of various categories as samples.

[0080] Optionally, after obtaining the equipotential star map, the equipotential star map can be input into a pre-trained first signal recognition model for initial recognition to obtain an initial signal category of the signal to be recognized. The first signal recognition model is trained using signal data of multiple categories as samples, and the initial signal category refers to which category the specific modulation type of the signal belongs to. These categories may include MQAM, BPSK, QPSK, 8PSK and MFSK, and the initial signal category is one of MQAM, BPSK, QPSK, 8PSK and MFSK.

[0081] Exemplarily, the first signal recognition model may be a convolutional neural network model, such as MobileNet V2. When training the first signal recognition model, both the training data and the verification data may include signal data of modulation categories such as MQAM, BPSK, QPSK, 8PSK, and MFSK.

[0082] S104, judging the signal category of the signal to be identified, if the signal category of the signal to be identified is multi-base quadrature amplitude modulation, determining the target signal type of the signal to be identified, and performing secondary identification on the signal to be identified based on a second signal identification model corresponding to the target signal type, to obtain an identification result of the signal to be identified.

[0083] The second signal recognition model is a MobileNet V2 model trained using signal data having a target signal type as samples.

[0084] It can be understood that when the signal to be identified is identified once in the above step S103, it can be determined whether the signal category of the signal to be identified is MQAM, BPSK, QPSK, 8PSK and MFSK. However, for the signal category, under the MQAM category, it is also necessary to determine the specific base of the orthogonal amplitude modulation of the signal to be identified, that is, to subdivide the modulation order. Therefore, after obtaining the signal category of the signal to be identified, the target signal type of the signal to be identified can be judged, so that the signal of the MQAM category can be identified twice to obtain the identification result of the signal to be identified.

[0085] Optionally, after obtaining the signal category of the signal to be identified, the signal category of the signal to be identified can be judged. If the signal category of the signal to be identified is MQAM, the signal to be identified can be further judged to determine the target signal type of the signal to be identified. The target signal type of the signal to be identified refers to which subcategory of the MQAM category the modulation type of the signal belongs to. The target signal type can include square quadrature amplitude modulation (square QAM for short) and cross quadrature amplitude modulation (cross QAM for short).

[0086] Optionally, after obtaining the target signal type of the signal to be identified, the signal to be identified can be re-identified based on a pre-trained second signal recognition model corresponding to the target signal type to obtain a recognition result of the signal to be identified. The second signal recognition model is trained using signal data having the target signal type as a sample, and the target signal type may include 16QAM, 32QAM, 64QAM, 128QAM, and 256QAM.

[0087] Exemplarily, for the second signal recognition model corresponding to the square QAM type, when the second signal recognition model corresponding to the square QAM type is trained, the training data and the verification data may include signal data of 16QAM, 64QAM and 256QAM types. Accordingly, when it is determined that the target signal type of the signal to be recognized is the square QAM type, the signal to be recognized can be secondary recognized based on the second signal recognition model corresponding to the square QAM type, so as to determine which type of 16QAM, 64QAM and 256QAM the signal to be recognized is.

[0088] Exemplarily, for the second signal recognition model corresponding to the cross-shaped QAM type, when the second signal recognition model corresponding to the cross-shaped QAM type is trained, the training data and the verification data may include signal data of 32QAM and 128QAM types. Accordingly, when it is determined that the target signal type of the signal to be recognized is a cross-shaped QAM type, the signal to be recognized can be secondary recognized based on the second signal recognition model corresponding to the cross-shaped QAM type, so as to determine which type of 32QAM and 128QAM the signal to be recognized is.

[0089] In this embodiment, by acquiring a signal to be identified, and generating an equipotential star diagram corresponding to the signal to be identified based on the signal to be identified, the symbol characteristics of the signal to be identified in different modulation modes can be characterized, so that the equipotential star diagram is input into the first signal recognition model for initial recognition, and the initial signal category of the signal to be identified is obtained, and the category of the signal to be identified is determined, and the initial signal category of the signal to be identified is obtained. On this basis, by judging the signal category of the signal to be identified, when the signal category of the signal to be identified is multi-level orthogonal amplitude modulation, the target signal type of the signal to be identified is determined, and based on the second signal recognition model corresponding to the target signal type, the signal to be identified is secondary identified, and the identification result of the signal to be identified is obtained. On the basis of the primary identification, the order of the signal to be identified in the orthogonal amplitude modulation can be determined, and the secondary identification of the multi-level orthogonal amplitude modulation can be realized. That is to say, when identifying signal modulation in the present application, first, the signal to be identified is identified once by a first signal recognition model obtained by training with signal data of multiple categories as samples, and the modulation method adopted by the signal to be identified is classified once to determine the category to which the modulation method of the signal to be identified belongs. Moreover, on the basis of the first recognition, the signal to be identified is secondarily identified by a second signal recognition model obtained by training with signal data of the target signal type as samples to determine the specific modulation method of the signal to be identified. Moreover, the signal data of multiple categories are different from the signal data of the target signal type, which can ensure the accuracy of signal modulation recognition and improve the generalization ability of signal recognition. In addition, due to the secondary recognition, the anti-interference ability of signal modulation recognition is also enhanced, thereby realizing the rapid and accurate recognition of digital modulated signals.

[0090] In a possible implementation, determining the target signal type of the signal to be identified in the above step S104 includes:

[0091] The target signal type of the signal to be identified is determined based on the high-order cumulative amount characteristics of the signal to be identified, and the target signal type includes: square quadrature amplitude modulation or cross quadrature amplitude modulation.

[0092] It can be understood that under the MQAM category, when the modulation order is an even power of 2, the distribution of the signal to be identified on the equipotential star map presents a square structure, and when the modulation order is an odd power of 2, the distribution of the signal to be identified on the equipotential star map presents a cross structure. Therefore, when it is determined that the modulation mode of the signal to be identified is a binary orthogonal amplitude modulation under the MQAM category, it can be determined whether the target signal type of the signal to be identified is square QAM or cross QAM based on the high-order cumulative amount characteristics of the signal to be identified.

[0093] Optionally, a high-order cumulant feature of the signal to be identified is calculated, and the obtained high-order cumulant feature is compared with a preset first high-order cumulant feature threshold, so as to determine whether the target signal type of the signal to be identified is square QAM or cross QAM.

[0094] Exemplarily, the preset first high-order cumulative feature threshold may be 0.9. If the high-order cumulative feature of the signal to be identified is greater than the preset first high-order cumulative feature threshold, the target signal type of the signal to be identified may be determined to be: square QAM. If not, the target signal type of the signal to be identified may be determined to be: cross QAM. The preset first high-order cumulative feature threshold may be configured in advance through code or may be obtained by user input.

[0095] Based on the high-order cumulative amount characteristics of the signal to be identified, the target signal type of the signal to be identified is determined, and the Gaussian white noise in the signal to be identified can be suppressed to avoid the interference caused by the Gaussian white noise in the signal to be identified, and effectively reduce the impact of noise on the target signal type. At the same time, the target signal type can be determined based on the high-order cumulative amount characteristics, and relatively stable performance can be maintained under different signal-to-noise ratio conditions, which can ensure the recognition accuracy and accuracy of the target signal type, and realize the effective classification of QAM with different modulation orders.

[0096] Figure 2 A schematic diagram of a flow chart for obtaining an identification result of a signal to be identified in the signal modulation identification method provided in an embodiment of the present application.

[0097] In one possible implementation, referring to Figure 2 As shown, in the above step S104, the second signal recognition model corresponding to the target signal type is used to perform secondary recognition on the signal to be recognized, and the recognition result of the signal to be recognized is obtained, including:

[0098] S201. Determine the normalized instantaneous amplitude feature of the signal to be identified.

[0099] It can be understood that after the target signal type is determined, the second signal recognition model corresponding to the target signal type can be used to perform secondary recognition on the signal to be recognized, thereby obtaining a recognition result of the signal to be recognized.

[0100] Optionally, before performing secondary recognition on the signal to be recognized based on the second signal recognition model corresponding to the target signal type, it is necessary to first calculate the normalized instantaneous amplitude characteristics of the signal to be recognized based on the signal to be recognized, so as to reduce the interference of noise in the signal to be recognized and eliminate the influence of channel gain and channel time-varying characteristics on the signal to be recognized.

[0101] S202: Input the normalized instantaneous amplitude feature into a second signal recognition model corresponding to the target signal type for secondary recognition to obtain a recognition result of the signal to be recognized.

[0102] Optionally, after the normalized instantaneous amplitude feature is obtained, the normalized instantaneous amplitude feature is input into a second signal recognition model corresponding to the target signal type for secondary recognition to obtain a recognition result of the signal to be recognized.

[0103] Exemplarily, for the second signal recognition model corresponding to the square QAM type, when the second signal recognition model corresponding to the square QAM type is trained, the training data and the verification data may include signal data of 16QAM, 64QAM and 256QAM types. Accordingly, when it is determined that the target signal type of the signal to be recognized is the square QAM type, the normalized instantaneous amplitude feature can be input into the second signal recognition model corresponding to the square QAM type for secondary recognition, thereby determining which type of 16QAM, 64QAM and 256QAM the signal to be recognized is.

[0104] Exemplarily, for the second signal recognition model corresponding to the cross-shaped QAM type, when the second signal recognition model corresponding to the cross-shaped QAM type is trained, the training data and the verification data may include signal data of 32QAM and 128QAM types. Accordingly, when it is determined that the target signal type of the signal to be recognized is a cross-shaped QAM type, the normalized instantaneous amplitude feature can be input into the second signal recognition model corresponding to the cross-shaped QAM type for secondary recognition, thereby determining which type of 32QAM and 128QAM the signal to be recognized is.

[0105] By determining the normalized instantaneous amplitude characteristics of the signal to be identified, the influence of noise on the signal to be identified can be reduced, and the robustness of signal modulation identification can be improved. The normalized instantaneous amplitude characteristics are input into the second signal recognition model corresponding to the target signal type for secondary recognition to obtain the recognition result of the signal to be identified. This can ensure the accuracy of signal modulation recognition and realize fast and accurate recognition of digital modulated signals.

[0106] The above describes in detail the processing procedure when the signal category of the signal to be identified is the MQAM category. It can be understood that the signal category of the signal to be identified can also be MFSK and other categories. The following describes the processing procedure when the signal category of the signal to be identified is MFSK and other categories.

[0107] In a possible implementation, the signal modulation identification method provided in the embodiment of the present application further includes:

[0108] If the signal category of the signal to be identified is MFSK, Fourier transform is performed on the signal to be identified to obtain the transformed signal to be identified, high-order cumulative quantity characteristics of the transformed signal to be identified are calculated, and the identification result of the signal to be identified is determined according to the high-order cumulative quantity characteristics.

[0109] It can be understood that when the signal to be identified is identified once in the above step S103, it can be determined whether the signal category of the signal to be identified is MQAM, BPSK, QPSK, 8PSK and MFSK. For the signal category, under the MFSK category, it is also necessary to subdivide the modulation order of the signal to be identified, that is, determine the base frequency shift keying of the signal to be identified. Therefore, after obtaining the signal category of the signal to be identified, the signal to be identified can be processed to determine whether the signal to be identified is binary frequency shift keying (2FSK for short) or quaternary frequency shift keying (4FSK for short) as the identification result of the signal to be identified.

[0110] Optionally, if the signal category of the signal to be identified is MFSK, the Fourier transform and the high-order cumulant feature can be applied jointly, and the Fourier transform is performed on the signal to be identified to obtain the transformed signal to be identified, and the high-order cumulant feature of the transformed signal to be identified is calculated to obtain the high-order cumulant feature parameter F3, and the high-order cumulant feature parameter F3 is compared with a preset second high-order cumulant feature threshold to determine the identification result of the signal to be identified.

[0111] Exemplarily, the preset second high-order cumulative feature threshold value may be 400. If the high-order cumulative feature parameter F3 is greater than the preset second high-order cumulative feature threshold value, the target signal type of the signal to be identified may be determined to be: 2FSK. If not, the target signal type of the signal to be identified may be determined to be: 4FSK. The preset second high-order cumulative feature threshold value may be configured in advance through code or may be obtained by user input.

[0112] When the signal category of the signal to be identified is MFSK, Fourier transform is performed on the signal to be identified, which can convert the MFSK category signal to be identified from the time domain to the frequency domain to obtain the transformed signal to be identified, which is helpful to reveal the hidden periodicity or main frequency that is not obvious in the signal to be identified, making the characteristics of the signal to be identified more conducive to extraction and analysis. On this basis, the high-order cumulative amount characteristics of the transformed signal to be identified are calculated, and the identification result of the signal to be identified is determined based on the high-order cumulative amount characteristics, thereby improving the accuracy and reliability of signal modulation identification.

[0113] In a possible implementation, the signal modulation identification method provided in the embodiment of the present application further includes:

[0114] If the signal category of the signal to be identified is a category other than MQAM and MFSK, the initial signal category is used as the identification result of the signal to be identified.

[0115] It can be understood that when the signal to be identified is identified once in the above step S103, it can be determined whether the signal category of the signal to be identified is MQAM, BPSK, QPSK, 8PSK and MFSK. For the signal category, under the BPSK, QPSK and 8PSK categories, the initial signal category is the identification result of the signal to be identified.

[0116] Optionally, if the signal category of the signal to be identified is a category other than MQAM and MFSK, the initial signal category is used as the identification result of the signal to be identified.

[0117] In a possible implementation, before generating an equipotential star map corresponding to the signal to be identified according to the signal to be identified in the above step S102, the following steps are included:

[0118] It is determined whether the signal to be identified is an in-phase orthogonal signal. If not, the signal to be identified is converted into an in-phase orthogonal signal.

[0119] It can be understood that after the signal to be identified is obtained, the data form of the signal to be identified may be an in-phase orthogonal signal (abbreviated as IQ signal or I / Q signal) or a non-in-phase orthogonal signal, such as real data. When generating an equipotential star map based on the signal to be identified, in order to ensure the accuracy of the generated result, it is necessary to enable the digital signal to be represented by a point on the complex plane, that is, it is necessary to ensure that the data form of the signal to be identified is in the form of an IQ signal.

[0120] Optionally, the signal to be identified is judged. If the signal to be identified is a 2×N or N×2 matrix, it is considered to be an in-phase orthogonal signal. If the signal to be identified is a 1×N or N×1 matrix, it is real data. When it is judged to be real data, the data form of the signal to be identified can be converted into an in-phase orthogonal signal through Hilbert transform.

[0121] By keeping the data form of the signal to be identified as an in-phase orthogonal signal, it helps to ensure that the characteristics and structure of the signal to be identified remain unchanged during the signal modulation identification process, thereby ensuring the accuracy of signal modulation identification.

[0122] Figure 3 A schematic diagram of a flow chart for obtaining an initial signal category of a signal to be identified in a signal modulation identification method provided in an embodiment of the present application.

[0123] In one possible implementation, referring to Figure 3 As shown, the above step S103 inputs the equipotential star map into the first signal recognition model for initial recognition to obtain the initial signal category of the signal to be recognized, including:

[0124] S301. Load a weight file of a first signal recognition model.

[0125] The weight file is used to indicate the parameters in the first signal recognition model.

[0126] Optionally, the first signal recognition model can be obtained by pre-training, and before the equipotential star map is input into the first signal recognition model, the weight file of the first signal recognition model obtained during training and the network structure of the first signal recognition model can be loaded first, wherein the weight file stores the parameter settings for indicating the first signal recognition model. The parameters may include the weight and bias parameters of each layer learned by the first signal recognition model during the training process.

[0127] S302: Load the weight file into the first signal recognition model.

[0128] Optionally, after obtaining the weight file, the weight file may be loaded into the network structure of the first signal recognition model to complete the setting of the first signal recognition model.

[0129] S303: Input the equipotential star map into the first signal recognition model for initial recognition to obtain an initial signal category of the signal to be recognized.

[0130] Optionally, after the weight file is loaded, the equipotential star map may be input into the loaded first signal recognition model to perform initial recognition, thereby obtaining an initial signal category of the signal to be recognized.

[0131] Loading the weight file of the first signal recognition model and loading the weight file into the first signal recognition model can save training time for the first signal recognition model and be used for initial recognition of the signal to be recognized more quickly, thereby inputting the equipotential planet map into the first signal recognition model for initial recognition, and quickly obtaining the initial signal category of the signal to be recognized. At the same time, by loading the weight file, the transfer learning ability of the first signal recognition model can also be improved, helping the first signal recognition model to converge faster in signal modulation recognition, and reducing the risk of overfitting, thereby ensuring the recognition performance of the first signal recognition model.

[0132] The above describes in detail the processing procedure of the signal modulation identification method provided in the embodiment of the present application. It can be understood that in the signal modulation identification method provided in the embodiment of the present application, multiple neural network models are involved, including a first signal recognition model and a second signal recognition model corresponding to the target signal type. The following is an illustrative description of the multiple neural network models involved in the signal modulation identification method provided in the embodiment of the present application.

[0133] It is worth noting that multiple neural network models such as the first signal recognition model and the second signal recognition model corresponding to the target signal type can be neural network models with the same network structure, but the sample data used in training and the weight files obtained are different, that is, when training the first signal recognition model, the sample data used can be signal data of multiple categories, which can specifically include: MQAM, BPSK, QPSK, 8PSK and MFSK, and the parameter settings corresponding to the first signal recognition model can be obtained as the weight file of the first signal recognition model. When training the second signal recognition model corresponding to the target signal type, different sample data can be used to obtain the second signal recognition model and weight file corresponding to the square QAM type, and the second signal recognition model and weight file corresponding to the cross QAM type.

[0134] Exemplarily, the neural network model used in the signal modulation identification method provided in the embodiment of the present application may be a convolutional neural network model, such as MobileNet V2.

[0135] Exemplarily, first, the overall structure of the MobileNet V2 model is illustrated. Table 1 below is a schematic diagram of the overall structure of the MobileNetV2 model:

[0136] Table 1 Overall network structure of MobileNet V2

[0137] enter layer t c n s 224×224×3 conv2d - 32 1 2 112×112×32 bottleneck 1 16 1 1 112×112×16 bottleneck 6 24 2 2 56×56×24 bottleneck 6 32 3 2 28×28×32 bottleneck 6 64 4 2 14×14×64 bottleneck 6 96 3 1 14×14×96 bottleneck 6 160 3 2 7×7×160 bottleneck 6 320 1 1 7×7×320 conv2d 1×1 - 1280 1 1 7×7×1280 avgpool 7×7 - - 1 - 1×1×1280 conv2d 1×1 - k - -

[0138] Among them, t is the multiple of the internal dimension increase of the botteneck layer, c is the number of channels, n is the number of times the botteneck layer is repeated, s is the step size, and k is the number of output categories, which can be determined by the number of signal modulation types.

[0139] For example, the following Table 2 shows the parameter settings of the MobileNet V2 model:

[0140] Table 2 Parameter settings of MobileNet V2 model

[0141] Model MobileNet V2 Batch size 64 Optimizer (op) Adam Learning rate adjustment strategy (lr) StepLR Regularization term weight factor (lamda) 0.001

[0142] For example, Figure 4 A schematic diagram of the structure of the bottleneck layer in the MobileNet V2 model in the signal modulation recognition method provided in an embodiment of the present application.

[0143] Exemplarily, the following takes the first signal recognition model as an example to illustrate the training process of the MobileNet V2 model and the process of obtaining the weight file.

[0144] Step S1 mainly includes data set preparation, using GNU Radio simulation software to generate signals with different modulation modes close to the real environment transmission, and drawing equipotential planet diagrams, dividing the training set, test set and validation set, specifically including:

[0145] Step S1-1, simulate the transmitter. Use a random source (abbreviated as RandomSource) in the GNU Radio simulation software, set the modulation mode (16QAM, 32QAM, 64QAM, 128QAM, 256QAM, BPSK, QPSK, 8PSK, 2FSK, 4FSK), the sampling rate is 64kHz, and the number of sampling points is 10000.

[0146] Step S1-2: Simulate the channel and use the channel module to realize signals with different signal-to-noise ratios (for example, 4-30 dB, with an interval of 2 dB) and multipath effects.

[0147] Step S1-3, simulate the receiving end. Use the channel module to simulate the asynchronous sampling clock, save the data after passing through the channel, a total of 10 (modulation modes) × 14 (signal-to-noise ratios) × 1000 (signals) = 140,000 signals, and draw an equipotential star map. For example, Figure 5 The equipotential star diagram of some signals in the signal modulation identification method provided in the embodiment of the present application is shown in FIG. Figure 5 As shown, the equipotential star diagram of the 16QAM modulation signal presents a square structure.

[0148] Step S1-4, making a signal sample set. The size of the equipotential star map in step S1-3 is adjusted to 224×224×3 as the input of the neural network model, and the characteristics of each digital modulation are intuitively displayed from the image, and the modulation mode is used as the label, that is, the output of the neural network model.

[0149] Step S1-5, divide the data set into the training set, the test set, and the validation set in a ratio of 3:1:1.

[0150] Step S2, model training. The training set in step S1 is used for training the neural network model, the features of the potential star map of each modulation mode are extracted, and used for self-learning of the neural network model, the weight file of the training phase is saved, and the parameters of the neural network model are optimized. Specifically including:

[0151] Step S2-1, initializing model parameters. Before training, the parameters of the neural network model are initialized.

[0152] Step S2-2, forward propagation: Input the training data into the neural network model, and after linear transformation of weights and biases and nonlinear processing of activation functions, the output result of the neural network model is obtained.

[0153] Step S2-3: Calculate the loss function. The loss function is used to evaluate the difference between the model's predicted value and the true label.

[0154] Step S2-4: Back propagation: Starting from the output layer, the gradient of the loss function to the parameters of each layer is calculated by the chain rule.

[0155] Step S2-5: Parameter update: Use an optimization method to update the parameters of the neural network model according to the gradient information.

[0156] Step S2-6, repeat training. Repeat steps S2-2 to S2-5 until a predetermined number of training rounds is reached or the loss function converges to a preset convergence value.

[0157] Step S2-7, save the weight file of the trained neural network model.

[0158] Step S3: Model evaluation and tuning. Use the test set in step S1 to evaluate the neural network model and perform neural network model tuning. Specifically, it includes:

[0159] Step S3-1, select evaluation indicators. For classification and recognition problems, you can use indicators such as accuracy, precision, recall, Fl-score, etc.

[0160] Step S3-2: Model evaluation: Using the validation set in step S1, calculate the indicators in step S3-1 to evaluate the performance and generalization ability of the neural network model.

[0161] Step S3-3: Model tuning: Adjust the model's hyperparameters based on the performance of the validation set, including: learning rate, regularization parameters, depth and width of the network structure, selection of activation function, etc.

[0162] Step S3-4: Final evaluation: Use the test set to conduct a final evaluation of the performance, robustness, and generalization ability of the neural network model.

[0163] Exemplarily, Table 3 below is a schematic table of recognition accuracy of the six modulation modes in the signal modulation recognition method provided in an embodiment of the present application under different signal-to-noise ratios. Referring to Table 3 below, taking QAM as a large category, the respective recognition accuracy rates can be obtained.

[0164] Table 3 Recognition accuracy of 6 modulation modes under different signal-to-noise ratios

[0165] Modulation 4 6 8 10 12 14 16 18 20 22 24 26 28 30 2FSK 99.08 100 100 100 100 100 100 100 100 100 100 100 100 100 4FSK 96.65 99.02 100 100 100 100 100 100 100 100 100 100 100 100 8PSK 94.79 92 99.5 99 100 99.5 100 99 100 100 100 100 100 99.5 BPSK 100 100 100 100 100 100 100 100 100 100 100 100 100 100 MQAM 99.5 98.5 100 99.5 100 100 100 100 100 100 100 100 100 100 QPSK 99.5 100 100 100 99.5 100 100 100 100 100 100 100 100 100

[0166] For example, Figure 6 This is a schematic diagram of recognition accuracy curves under different signal-to-noise ratios for six modulation modes in the signal modulation recognition method provided in an embodiment of the present application, Figure 7 A schematic diagram of confusion matrices of six modulation modes in the signal modulation identification method provided in an embodiment of the present application.

[0167] Exemplarily, the following Table 4 is a schematic table of the recognition results of the normalized instantaneous amplitude of the square QAM in the signal modulation recognition method provided in the embodiment of the present application in the neural network. Referring to the following Table 4, the respective recognition results can be obtained.

[0168] Table 4 Recognition results of normalized instantaneous amplitude of square QAM in neural network

[0169]

[0170] For example, Figure 8 A schematic diagram of the recognition accuracy curve under different signal-to-noise ratios of the square QAM in the signal modulation recognition method provided in an embodiment of the present application, wherein the horizontal axis is: signal-to-noise ratio, and the vertical axis is: recognition accuracy. Fig. 9 A schematic diagram of the confusion matrix of square QAM in the signal modulation identification method provided in an embodiment of the present application.

[0171] Exemplarily, the following Table 5 is a schematic table of the recognition results of the normalized instantaneous amplitude of the cross QAM in the signal modulation recognition method provided in the embodiment of the present application in the neural network. Referring to the following Table (5), the respective recognition results can be obtained.

[0172] Table 5 Recognition results of normalized instantaneous amplitude of cross QAM in neural network

[0173]

[0174] For example, Fig.10 This is a schematic diagram of the recognition accuracy curve under different signal-to-noise ratios of the cross QAM in the signal modulation recognition method provided in an embodiment of the present application, wherein the horizontal axis is: signal-to-noise ratio, and the vertical axis is: recognition accuracy. Fig.11 A schematic diagram of a confusion matrix of a cross QAM in a signal modulation identification method provided in an embodiment of the present application.

[0175] Based on the same inventive concept, a signal modulation identification device corresponding to the signal modulation identification method is also provided in the embodiment of the present application. Since the principle of solving the problem by the device in the embodiment of the present application is similar to the above-mentioned signal modulation identification method in the embodiment of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0176] Fig.12 A schematic diagram of a signal modulation identification device provided in an embodiment of the present application, referring to Fig.12 As shown, the device includes: an acquisition module 1201, a generation module 1202, a primary recognition module 1203 and a secondary recognition module 1204;

[0177] An acquisition module 1201 is used to acquire a signal to be identified;

[0178] A generating module 1202 is used to generate an equipotential star map corresponding to the signal to be identified according to the signal to be identified;

[0179] The initial recognition module 1203 is used to input the equipotential star map into the first signal recognition model for initial recognition to obtain the initial signal category of the signal to be recognized. The first signal recognition model is a MobileNet V2 model trained with signal data of multiple categories as samples;

[0180] The secondary recognition module 1204 is used to determine the signal category of the signal to be recognized. If the signal category of the signal to be recognized is multi-base quadrature amplitude modulation, the target signal type of the signal to be recognized is determined, and the signal to be recognized is secondary recognized based on a second signal recognition model corresponding to the target signal type to obtain a recognition result of the signal to be recognized. The second signal recognition model is a MobileNet V2 model trained using signal data with the target signal type as samples.

[0181] In a possible implementation, the secondary identification module 1204 is specifically configured to:

[0182] The target signal type of the signal to be identified is determined based on the high-order cumulative amount characteristics of the signal to be identified, and the target signal type includes: square quadrature amplitude modulation or cross quadrature amplitude modulation.

[0183] In a possible implementation, the secondary identification module 1204 is specifically configured to:

[0184] Determine the normalized instantaneous amplitude characteristics of the signal to be identified;

[0185] The normalized instantaneous amplitude feature is input into the second signal recognition model corresponding to the target signal type for secondary recognition to obtain the recognition result of the signal to be recognized.

[0186] In a possible implementation, the apparatus further includes: a determination module;

[0187] The determination module is used to perform Fourier transform on the signal to be identified if the signal category of the signal to be identified is MFSK, obtain the transformed signal to be identified, calculate the high-order cumulative amount characteristics of the transformed signal to be identified, and determine the identification result of the signal to be identified based on the high-order cumulative amount characteristics.

[0188] In a possible implementation, the determining module is further configured to:

[0189] If the signal category of the signal to be identified is a category other than MQAM and MFSK, the initial signal category is used as the identification result of the signal to be identified.

[0190] In a possible implementation, the generating module 1202 is further configured to:

[0191] It is determined whether the signal to be identified is an in-phase orthogonal signal. If not, the signal to be identified is converted into an in-phase orthogonal signal.

[0192] In a possible implementation, the initial identification module 1203 is specifically configured to:

[0193] Loading a weight file of the first signal recognition model, where the weight file is used to indicate parameters in the first signal recognition model;

[0194] Loading the weight file into the first signal recognition model;

[0195] The equipotential star map is input into the first signal recognition model for initial recognition to obtain the initial signal category of the signal to be recognized.

[0196] For descriptions of the processing flow of each module in the device and the interaction flow between each module, reference may be made to the relevant descriptions in the above method embodiment, which will not be described in detail here.

[0197] The present application embodiment also provides an electronic device 1300, such as Fig.13 As shown, Fig.13 The structural diagram of the electronic device provided in the embodiment of the present application includes: a processor 1301, a memory 1302, and optionally, a bus 1303. The memory 1302 stores machine-readable instructions executable by the processor 1301 (for example, Fig.12 In the device, the acquisition module 1201, the generation module 1202, the initial recognition module 1203 and the execution instructions corresponding to the secondary recognition module 1204 are obtained, and when the electronic device 1300 is running, the processor 1301 communicates with the memory 1302 through the bus 1303, and when the machine-readable instructions are executed by the processor 1301, the steps of the above-mentioned signal modulation recognition method are performed.

[0198] An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned signal modulation identification method are executed.

[0199] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the method embodiment, and will not be repeated in this application. In the several embodiments provided in this application, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0200] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or part of the technical solution that contributes to the prior art or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, RandomAccess Memory), disk or optical disk and other media that can store program code.

[0201] The above are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be covered by the protection scope of the present application.

Claims

1. A signal modulation identification method, characterized in that: include: Acquire a signal to be identified; According to the signal to be identified, generating an equipotential planet diagram corresponding to the signal to be identified; Inputting the equipotential star map into a first signal recognition model for initial recognition to obtain an initial signal category of the signal to be recognized, wherein the first signal recognition model is a MobileNet V2 model trained using signal data of multiple categories as samples; The signal category of the signal to be identified is judged. If the signal category of the signal to be identified is multi-base quadrature amplitude modulation, the target signal type of the signal to be identified is determined, and the signal to be identified is re-identified based on a second signal recognition model corresponding to the target signal type to obtain a recognition result of the signal to be identified, where the second signal recognition model is a MobileNetV2 model trained with signal data having the target signal type as a sample; The determining the target signal type of the signal to be identified includes: Determine the target signal type of the signal to be identified based on the high-order cumulative amount characteristics of the signal to be identified, the target signal type includes: square quadrature amplitude modulation or cross quadrature amplitude modulation; The performing secondary recognition on the signal to be recognized based on the second signal recognition model corresponding to the target signal type to obtain a recognition result of the signal to be recognized includes: Determining a normalized instantaneous amplitude characteristic of the signal to be identified; Inputting the normalized instantaneous amplitude feature into a second signal recognition model corresponding to the target signal type for secondary recognition to obtain a recognition result of the signal to be recognized; The method further comprises: If the signal type of the signal to be identified is multi-level frequency shift keying, Fourier transform is performed on the signal to be identified to obtain a transformed signal to be identified, high-order cumulative quantity characteristics of the transformed signal to be identified are calculated, and an identification result of the signal to be identified is determined according to the high-order cumulative quantity characteristics; The method further comprises: If the signal category of the signal to be identified is a category other than multi-level quadrature amplitude modulation and multi-level frequency shift keying, the initial signal category is used as the identification result of the signal to be identified.

2. The signal modulation identification method according to claim 1, characterized in that: Before generating an equipotential star map corresponding to the signal to be identified according to the signal to be identified, the method includes: It is determined whether the signal to be identified is an in-phase orthogonal signal; if not, the signal to be identified is converted into an in-phase orthogonal signal.

3. The signal modulation identification method according to claim 1 or 2, characterized in that: The step of inputting the equipotential star map into a first signal recognition model for initial recognition to obtain an initial signal category of the signal to be recognized includes: Loading a weight file of the first signal recognition model, wherein the weight file is used to indicate parameters in the first signal recognition model; Loading the weight file into the first signal recognition model; The equipotential star map is input into the first signal recognition model for initial recognition to obtain an initial signal category of the signal to be recognized.

4. A signal modulation identification device, characterized in that: include: An acquisition module, used for acquiring a signal to be identified; A generating module, used for generating an equipotential star map corresponding to the signal to be identified according to the signal to be identified; A primary recognition module, used for inputting the equipotential star map into a first signal recognition model for primary recognition to obtain an initial signal category of the signal to be recognized, wherein the first signal recognition model is a MobileNet V2 model trained with signal data of multiple categories as samples; A secondary recognition module, used to judge the signal category of the signal to be recognized, and if the signal category of the signal to be recognized is multi-base quadrature amplitude modulation, determine the target signal type of the signal to be recognized, and perform secondary recognition on the signal to be recognized based on a second signal recognition model corresponding to the target signal type to obtain a recognition result of the signal to be recognized, wherein the second signal recognition model is a MobileNet V2 model trained with signal data having the target signal type as a sample; The secondary identification module is specifically used for: Determine the target signal type of the signal to be identified based on the high-order cumulative amount characteristics of the signal to be identified, the target signal type includes: square quadrature amplitude modulation or cross quadrature amplitude modulation; The secondary identification module is specifically used for: Determining a normalized instantaneous amplitude characteristic of the signal to be identified; Inputting the normalized instantaneous amplitude feature into a second signal recognition model corresponding to the target signal type for secondary recognition to obtain a recognition result of the signal to be recognized; The device further comprises: a determination module; The determination module is used for, if the signal type of the signal to be identified is multi-level frequency shift keying, performing Fourier transform on the signal to be identified to obtain a transformed signal to be identified, calculating a high-order cumulative amount feature of the transformed signal to be identified, and determining a recognition result of the signal to be identified according to the high-order cumulative amount feature; The determining module is further used for: If the signal category of the signal to be identified is a category other than multi-level quadrature amplitude modulation and multi-level frequency shift keying, the initial signal category is used as the identification result of the signal to be identified.

5. An electronic device, characterized in that: include: A processor and a memory, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor executes the machine-readable instructions to perform the steps of the signal modulation identification method as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the signal modulation identification method as claimed in any one of claims 1 to 3 are executed.

Images